I’m a huge fan of iNaturalist — I use it all the time for my own interests, and I’ve also incorporated it into an assignment in introductory biology. Students are all walking around with cameras in their phones, so I have them create an iNaturalist account and find some living thing in their environment, take a picture, and report back with an accurate Latin binomial. Anything goes — take a photo of a houseplant in their dorm room, a squirrel on the campus mall, a bug on a leaf, whatever. The nice thing about iNaturalist is that even if you don’t know, the software will attempt an automatic recognition, and you’ll get community feedback and eventually get a good identification. It has a huge userbase, and one of its virtues is that there always experts who can help you get an answer.
Basically, iNaturalist already has a kind of distributed human intelligence, so why would they want an artificial intelligence bumbling about, inserting hallucinations into the identifications? The answer is they shouldn’t. But now they’ve got one, thanks to a $1.5 million grant from Google. It’s advantageous to Google, because it gives them another huge database of human-generated data to plunder, but the gain for humans and other naturalists is non-existent.
On June 10 the nonprofit organization iNaturalist, which runs a popular online platform for nature observers, announced in a blog post that it had received a $1.5-million grant from Google.org Accelerator: Generative AI—an initiative of Google’s philanthropic arm—to “help build tools to improve the identification experience for the iNaturalist community.” More than 3.7 million people around the world—from weekend naturalists to professional taxonomists—use the platform to record observations of wild organisms and get help with identifying the species. To date, the iNaturalist community has logged upward of 250 million observations of more than half a million species, with some 430,000 members working to identify species from photographs, audio and text uploaded to the database. The announcement did not go over well with iNaturalist users, who took to the comments section of the blog post and a related forum, as well as Bluesky, in droves to voice their concerns.
Currently, the identification experience is near perfect. How will Google improve it? They should be working on improving the user experience on their search engine, which has become a trash heap of AI slop, rather than injecting more AI slop into the iNaturalist experience. The director of iNaturalist is trying to save face by declaring that this grant to insert generative AI into iNaturalist will not be inserting generative AI into iNaturalist, when that’s the whole reason for Google giving them the grant.
I can assure you that I and the entire iNat team hates the AI slop that’s taking over the internet as much as you do.
… there’s no way we’re going to unleash AI generated slop onto the site.
Here’s a nice response to that.
Those are nice words, but AI-generated slop is still explicitly the plan. iNaturalist’s grant deliverable is “to have an initial demo available for select user testing by the end of 2025.”
You can tell what happened — Google promised iNaturalist free money if they would just do something, anything, that had some generative AI in it. iNaturalist forgot why people contribute at all, and took the cash.
The iNaturalist charity is currently “working on a response that should answer most of the major questions people have and provide more clarity.”
They’re sure the people who do the work for free hate this whole plan only because there’s not enough “clarity” — and not because it’s a terrible idea.
People are leaving iNaturalist over this bad decision. The strength of iNaturalist has always been the good, dedicated people who work so hard at it, so any decision that drives people away and replaces them with a hallucinating bot is a bad decision.
I can really recommend ObsIdentify, but I guess it’s mostly for Europe.
I love iNaturalist. I moved to Malaysia a few years ago and photograph the wildlife vertebrate and invertebrate that lurks around my house. I post them on iNaturalist with my attempt at identification and usually get several suggestions within a day. Not only am I learning about the local wildlife I get to know who ese is out there and interested. Lately I have been photographing some interesting fungi which I know nothing about. I’ll see how that goes. As for AI, I did some work on testing LLMs for AI a while back. The model did not approve when I corrected its stupidity. It cant even use Google properly.
I have as much distaste for AI as the next guy, but this whole issue is getting overblown. The developers have clarified that no generative AI will be used and that Google is not receiving any data in exchange. Also, AI is already used and integrated into iNat to provide species suggestions, would you advocate removing this feature too simply because it is AI?
Well, that’s proactive of users, to leave before the initial demo.
Also, nothing totally lacks downsides: https://www.ccwa.org.au/good_intentions_gone_wrong
With the note that i have never used iNaturalist: there are probably machine learning approaches iNaturalist could use to improve the experience, but the fact they’re billing this as generative “AI” suggests they have no interest in that sort of approach. This seems like Google buying PR, and iNaturalist should have held out for more money
Haven’t used it myself but I know iNaturalist is widely used by many people locally who do volunteer bushcare and bushwalking and are intrested in our local flora and fauna here. My brother is a big fan of it for one as are many other friends. So, yeah, this is a big worry and hope they don’t mess it up and make it unreliable.
Good textbooks and personal knowledge from those that have taken the time to learn their wildflowers, insects, fungi etc are still so vital and irreplacebale really.
For all the good they offer, computers and phones and the web still never 100% trustworthy or reliable then there’s battery levels, technical isssues – and, ofc, the nuisance that is botantists changing plant species names all the flippin’ time.. Grrr.. (I know, they have their reasons but.still. )
@dangerousbeans #4:
That’s my thinking as well. I use a couple of apps for identifying fauna that I assume rely on machine learning. Very useful as you get an instant answer, usually with a few alternatives and a “percent match” figure. But the moment I hear AI I’m thinking scam, slop or both.
Well, there’s still GBIF. Can’t really trust these private actors now can we?
@5 StevoR:
“Good textbooks and personal knowledge from those that have taken the time to learn their wildflowers, insects, fungi etc are still so vital and irreplacebale really.”
ideally those are the people using iNat to identify the millions of observations, there are lots of experts on there who specialize in certain groups.
… until the platform ruins its reputation by joining the genAI slop fray and drives those people away.
I still remember my first iNaturalist ID: I took a photo of a large and scary-looking insect from a distance, cropped out everything but a small square containing that (rather blurry and low-res) insect, and got back “Eastern cicada hunter-killer wasp”. I’m pretty sure no human was directly involved with making that ID (see Anthony Rozewski #3 about already-integrated species-recognition AI).
I’m also pretty sure that the ID was correct — cicada hunter-killers are large and scary-looking ground-nesters (and the subject of my photo had nested in the ground). I was relieved, though, to learn that they are not aggressive, and their sting is not particularly painful, and the males cannot sting.
This was several years ago, and I didn’t really use the app a whole lot more.
@10 Owlmirror
the ‘computer vision’ model is AI but not generative AI. it makes suggestions based on a vast knowledge base made up of previous confirmed IDs on iNat. the wasp you uploaded and IDed would not be included into the computer vision’s data set until an agreeing ID is added (by a human).
for charismatic organisms the CV works very well, though it tends to be overly-confident (suggesting species-level identifications for things that can’t reliably be identified by a photo). a lot of the IDs i submit are taking the observation up a level or two – adjusting from species to genus or family. it’s the human IDers that make iNat what it is, and a lot of them would be happy with the computer vision model being reined in a bit – there is a feedback loop in which iNat IDers ‘learn’ a species based on the CV and confirm incorrect IDs, teaching the CV that it is correct when it isn’t.
the newer concept seems to rely more on user comments and [???] sources to explain ‘why’ the CV selects what it does (when in reality it is just matching pixels). a lot of people in the iNat community have been pushing for user-curated wiki pages for years, and it seems foolish to ignore the large group of volunteer experts in favor of a generative AI solution.
i think a lot of the backlash is the association with google specifically. i know a lot of recurring donors have ended their monthly donations since ‘you don’t need my money anymore.’ folks don’t really trust google all that much, given… well… google.
@9. coffeepott :
Well, yeah, I guess them too although I’m thinking more the people who learnt the hard and old way years ago as I did with various textbooks such as Its Blue With Five Petals* and doing courses like Native Plant ID at TAFE or in field workshops or from the old-fashioned field pocket-books and botany guide brochures and non-computer, non-phone (ironically?) dead tree tech!
From listening to and walking & working with – with textbooks and hand lens in backpacks with us – older and more experienced botanists and people like the plant ID guru and native orchid guided walk leader we have in our local volunteer bushcare group who are “amateur” (in designation anyhow – although at least as knowledgeable as some – even many? – professionals in my view.) botanists / entomologists / mycologists, etc .. Old fahsioned but reckon there’s still a lot to be said for that rather than computer and phone knowledge..Guess that shows my age / mindset?
.* See among other places : https://www.itsbluewithfivepetals.com.au/
those are the people i’m referring to as experts. i’ve spent enough time keying out arthropods to know to most avoid adding species-level IDs on iNat.
occasionally IDs on iNat will be accompanied with a quote from a species description/etc or at least a reference. not occasionally enough, of course! iNat does allow users to add a comment, tag another user, and ask how they came to an ID.
there is a very high percentage of observations that are crummy cell phone photos that shouldn’t’ve been uploaded, however! the superior arthropod website for north americans, bugguide, has a nice system of rejecting useless photos.